Re: Index management

"Noons" <wizofoz2k_at_yahoo.com.au> wrote in message news:40979aeb$0$4547$afc38c87_at_news.optusnet.com.au...
> Well, I think the thread is saying very clearly that index-level
clustering
> and other "defrag" techniques are not that important as soon as multi-user
> is involved. Which is the vast majority of the time. Should we really
bother
> with the odd "other" conditions? Does it really gain us much anymore?
> </brain-storming>

Hi Nuno,

Need to be a little careful here.

Any techniques that can group or cluster like data together so that we can read this data by accessing fewer blocks is a good thing providing the benefits we gain are not counter-balanced by the costs of maintaining such structures.

So for example as I mention in my presentation, if by re-ordering the data in a table to match the order of our most "important" index via business critical index range scan operations, we can dramatically reduce the number of different data blocks we need to visit, which could then noticeably improves important business response times, then that's a good thing. Same potentially goes for some clustered data structures. Like I said, it all depends on the benefits vs. the maintenance costs.

Cheers

Richard Received on Tue May 04 2004 - 08:48:03 CDT